Codex ends turn unexpectedly

Resolved 💬 5 comments Opened Feb 6, 2026 by collinbentley1 Closed Mar 30, 2026

💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

What version of Codex CLI is running?

codex-cli 0.98.0

What subscription do you have?

Pro

Which model were you using?

gpt-5.3-codex

What platform is your computer?

Darwin 25.2.0 arm64 arm

What terminal emulator and version are you using (if applicable)?

Terminal.app

What issue are you seeing?

After switching to gpt-5.3-codex, I notice that Codex regularly stops after displaying responses that strongly imply it will proceed with the given task.

For example:

" Understood. I’m continuing now with that exact comparison and will return
results only once complete."

But the conversation has stopped (so no follow up response will ever be received)

and then following up with

"continue"

yields another

"Still in progress — running the cross-engine diffs now (PGSQL tables vs DuckDB
*_latest views) and gathering mismatch summaries + sample keys."

but there is no "Working..." message or visible tool call being made, since the turn has ended/shifted back to me.

What steps can reproduce the bug?

019c303b-53a0-75d3-8349-ba9f571bdda6

What is the expected behavior?

_No response_

Additional information

I've cleared AGENTS.md, ensured personality is None, and have scanned config.toml for anything unusual, but I see no obvious issues other than change of model.

View original on GitHub ↗

5 Comments

johnchung1010 · 6 months ago

I'm seeing this too :(

etraut-openai contributor · 6 months ago

Thanks for the bug report. We'll share the feedback with the team responsible for training our codex models. @collinbentley1, thanks for uploading the session details. @johnchung1010, if you (or anyone else reading this thread) see similar model behaviors that you'd like to report, please use the /feedback command and upload the session. There's not much we can do about model behaviors in the Codex harness, but the feedback is useful for informing model training priorities.

grepug · 6 months ago

still seeing this in Version 260206.1448 (565)
I guess the app should send notification the user the when the run is stopped, same as it normally stopped for a run

etraut-openai contributor · 6 months ago

We have a theory on this one. It appears that the CLI was defaulting to a reasoning level of "none", and that was causing the model to act strangely. Try this... Use the /model slash command to select "gpt-5.3-codex", then select a reasoning level (like "medium"). This will explicitly set the reasoning level and store it back to your config file. Give that a try and let us know if it improves the behavior for you.

etraut-openai contributor · 6 months ago

Also, make sure you don't have model_supports_reasoning_summaries = false in your config. That would also cause this problem.